@simonw

–

18 April 2024 17h18

How does that do Claude tokenization? I didn't think they'd released code for that

Base Knowledge Requirements for AI Summarization Tasks

By

–

18 April 2024 17h17

A question I've had for a while is how much base knowledge a model needs in order to be good at summarization etc Does it need to know that an elephant is larger try an a mouse (or what an elephant and a mouse are) in order to summarize a meeting agenda?

LLM Document Analysis Task Scoring Methodology

By

–

18 April 2024 17h09

That's a different issue for me – I want scores for the tasks that are most useful for document analysis with LLMs

Best LLM Benchmarks for Summarization and RAG Tasks

By

–

18 April 2024 17h06

LLM benchmark question: benchmarks like MMLU do a lot of testing for knowledge – what are the most interesting benchmarks for if I don't care as much about what the model "knows" but more about how good it is at tasks like summarization, data extraction and RAG Q&A against input?

DSPy Requires Advanced Prompt Engineering Expertise

By

–

18 April 2024 14h14

My argument is that you have to already be a top level expert in prompt engineering to use something like DSPy

Promising AI Approach Gains Traction After Public Discussion

By

–

18 April 2024 14h09

I enjoyed the HN conversation about it – looks promising to me https://
news.ycombinator.com/item?id=400676
77
… Love the clarity of the write-up. I expect we will see if the approach has legs within a month or two now that the information is out there

Running AI Models on GPU: Learning Technical Implementation

By

–

18 April 2024 13h46

The point of the messing around is to learn more about how to run models on a GPU

Mixtral 8x22B Now Available via Mistral AI Platform API

By

–

18 April 2024 5h15

Mixtral 8x22B is now available via the @MistralAILabs La Platforme API If you previously installed the llm-mistral plugin run "llm mistral refresh" to refresh the list of available models – otherwise a fresh install will provide it Released 0.3.1 anyway: https://
github.com/simonw/llm-mis
tral/releases/tag/0.3.1
…

Open Source Models Prompt Injection Security Flaw

By

–

18 April 2024 4h46

Yeah, I wouldn't trust that! That does highlight interesting flaw in a lot of open models though: I think there are some models that use strings like [INST] without even reserving a token for them, which opens up all sorts of additional potential prompt injection mischief